Abstract: During deformation monitoring by a measuring robot， the measurement signal drifts due to meteorological factors， resulting in significant measurement errors. If raw data are directly exported as final monitoring results， this will lead to wide deformation fitting intervals and low monitoring accuracy. To address this issue， this paper proposes a method for deformation monitoring in complex tunnel construction environments using measuring robots. This method utilizes measuring robots to conduct high-precision measurements and effectively eliminates errors induced by meteorological factors through differential correction technology. The differentially corrected measurement values serve as the observation input for the Kalman filtering algorithm to dynamically estimate and predict the deformation states of the tunnel. Data containing tunnel deformation predictions are taken as feature vectors and combined with a Long Short-Term Memory （LSTM） network to calculate dynamic early warning risk values. The test results showed that the proposed method achieved a significantly narrower deformation fitting interval and more reliable monitoring accuracy.